2 * Copyright (c) 1982, 1986, 1989, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
34 * $FreeBSD: src/sys/kern/kern_proc.c,v 1.63.2.9 2003/05/08 07:47:16 kbyanc Exp $
35 * $DragonFly: src/sys/kern/kern_proc.c,v 1.45 2008/06/12 23:25:02 dillon Exp $
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/sysctl.h>
42 #include <sys/malloc.h>
45 #include <sys/filedesc.h>
47 #include <sys/dsched.h>
48 #include <sys/signalvar.h>
49 #include <sys/spinlock.h>
53 #include <vm/vm_map.h>
55 #include <machine/smp.h>
57 #include <sys/spinlock2.h>
58 #include <sys/mplock2.h>
60 static MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
61 MALLOC_DEFINE(M_SESSION, "session", "session header");
62 MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
63 MALLOC_DEFINE(M_LWP, "lwp", "lwp structures");
64 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
66 int ps_showallprocs = 1;
67 static int ps_showallthreads = 1;
68 SYSCTL_INT(_security, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
70 "Unprivileged processes can see proccesses with different UID/GID");
71 SYSCTL_INT(_security, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
72 &ps_showallthreads, 0,
73 "Unprivileged processes can see kernel threads");
75 static void pgdelete(struct pgrp *);
76 static void orphanpg(struct pgrp *pg);
77 static pid_t proc_getnewpid_locked(int random_offset);
82 struct pidhashhead *pidhashtbl;
84 struct pgrphashhead *pgrphashtbl;
86 struct proclist allproc;
87 struct proclist zombproc;
88 struct spinlock allproc_spin;
91 * Random component to nextpid generation. We mix in a random factor to make
92 * it a little harder to predict. We sanity check the modulus value to avoid
93 * doing it in critical paths. Don't let it be too small or we pointlessly
94 * waste randomness entropy, and don't let it be impossibly large. Using a
95 * modulus that is too big causes a LOT more process table scans and slows
96 * down fork processing as the pidchecked caching is defeated.
98 static int randompid = 0;
101 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
106 error = sysctl_handle_int(oidp, &pid, 0, req);
107 if (error || !req->newptr)
109 if (pid < 0 || pid > PID_MAX - 100) /* out of range */
111 else if (pid < 2) /* NOP */
113 else if (pid < 100) /* Make it reasonable */
119 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
120 0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
123 * Initialize global process hashing structures.
129 LIST_INIT(&zombproc);
130 spin_init(&allproc_spin);
132 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
133 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
138 * Is p an inferior of the current process?
141 inferior(struct proc *p)
143 for (; p != curproc; p = p->p_pptr)
150 * Locate a process by number
157 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
165 * Locate a process group by number
172 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
173 if (pgrp->pg_id == pgid)
180 * Move p to a new or existing process group (and session)
183 enterpgrp(struct proc *p, pid_t pgid, int mksess)
185 struct pgrp *pgrp = pgfind(pgid);
187 KASSERT(pgrp == NULL || !mksess,
188 ("enterpgrp: setsid into non-empty pgrp"));
189 KASSERT(!SESS_LEADER(p),
190 ("enterpgrp: session leader attempted setpgrp"));
193 pid_t savepid = p->p_pid;
198 KASSERT(p->p_pid == pgid,
199 ("enterpgrp: new pgrp and pid != pgid"));
200 if ((np = pfind(savepid)) == NULL || np != p)
202 MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP,
205 struct session *sess;
210 MALLOC(sess, struct session *, sizeof(struct session),
211 M_SESSION, M_WAITOK);
213 sess->s_sid = p->p_pid;
215 sess->s_ttyvp = NULL;
217 bcopy(p->p_session->s_login, sess->s_login,
218 sizeof(sess->s_login));
219 p->p_flag &= ~P_CONTROLT;
220 pgrp->pg_session = sess;
221 KASSERT(p == curproc,
222 ("enterpgrp: mksession and p != curproc"));
224 pgrp->pg_session = p->p_session;
225 sess_hold(pgrp->pg_session);
228 LIST_INIT(&pgrp->pg_members);
229 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
231 SLIST_INIT(&pgrp->pg_sigiolst);
232 lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
233 } else if (pgrp == p->p_pgrp)
237 * Adjust eligibility of affected pgrps to participate in job control.
238 * Increment eligibility counts before decrementing, otherwise we
239 * could reach 0 spuriously during the first call.
242 fixjobc(p, p->p_pgrp, 0);
244 LIST_REMOVE(p, p_pglist);
245 if (LIST_EMPTY(&p->p_pgrp->pg_members))
248 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
253 * remove process from process group
256 leavepgrp(struct proc *p)
259 LIST_REMOVE(p, p_pglist);
260 if (LIST_EMPTY(&p->p_pgrp->pg_members))
267 * delete a process group
270 pgdelete(struct pgrp *pgrp)
274 * Reset any sigio structures pointing to us as a result of
275 * F_SETOWN with our pgid.
277 funsetownlst(&pgrp->pg_sigiolst);
279 if (pgrp->pg_session->s_ttyp != NULL &&
280 pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
281 pgrp->pg_session->s_ttyp->t_pgrp = NULL;
282 LIST_REMOVE(pgrp, pg_hash);
283 sess_rele(pgrp->pg_session);
288 * Adjust the ref count on a session structure. When the ref count falls to
289 * zero the tty is disassociated from the session and the session structure
290 * is freed. Note that tty assocation is not itself ref-counted.
293 sess_hold(struct session *sp)
299 sess_rele(struct session *sp)
301 KKASSERT(sp->s_count > 0);
302 if (--sp->s_count == 0) {
303 if (sp->s_ttyp && sp->s_ttyp->t_session) {
304 #ifdef TTY_DO_FULL_CLOSE
305 /* FULL CLOSE, see ttyclearsession() */
306 KKASSERT(sp->s_ttyp->t_session == sp);
307 sp->s_ttyp->t_session = NULL;
309 /* HALF CLOSE, see ttyclearsession() */
310 if (sp->s_ttyp->t_session == sp)
311 sp->s_ttyp->t_session = NULL;
314 kfree(sp, M_SESSION);
319 * Adjust pgrp jobc counters when specified process changes process group.
320 * We count the number of processes in each process group that "qualify"
321 * the group for terminal job control (those with a parent in a different
322 * process group of the same session). If that count reaches zero, the
323 * process group becomes orphaned. Check both the specified process'
324 * process group and that of its children.
325 * entering == 0 => p is leaving specified group.
326 * entering == 1 => p is entering specified group.
329 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
331 struct pgrp *hispgrp;
332 struct session *mysession = pgrp->pg_session;
335 * Check p's parent to see whether p qualifies its own process
336 * group; if so, adjust count for p's process group.
338 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
339 hispgrp->pg_session == mysession) {
342 else if (--pgrp->pg_jobc == 0)
347 * Check this process' children to see whether they qualify
348 * their process groups; if so, adjust counts for children's
351 LIST_FOREACH(p, &p->p_children, p_sibling)
352 if ((hispgrp = p->p_pgrp) != pgrp &&
353 hispgrp->pg_session == mysession &&
354 p->p_stat != SZOMB) {
357 else if (--hispgrp->pg_jobc == 0)
363 * A process group has become orphaned;
364 * if there are any stopped processes in the group,
365 * hang-up all process in that group.
368 orphanpg(struct pgrp *pg)
372 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
373 if (p->p_stat == SSTOP) {
374 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
384 * Add a new process to the allproc list and the PID hash. This
385 * also assigns a pid to the new process.
387 * MPALMOSTSAFE - acquires mplock for karc4random() call
390 proc_add_allproc(struct proc *p)
394 if ((random_offset = randompid) != 0) {
396 random_offset = karc4random() % random_offset;
400 spin_lock_wr(&allproc_spin);
401 p->p_pid = proc_getnewpid_locked(random_offset);
402 LIST_INSERT_HEAD(&allproc, p, p_list);
403 LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
404 spin_unlock_wr(&allproc_spin);
408 * Calculate a new process pid. This function is integrated into
409 * proc_add_allproc() to guarentee that the new pid is not reused before
410 * the new process can be added to the allproc list.
412 * MPSAFE - must be called with allproc_spin held.
416 proc_getnewpid_locked(int random_offset)
418 static pid_t nextpid;
419 static pid_t pidchecked;
423 * Find an unused process ID. We remember a range of unused IDs
424 * ready to use (from nextpid+1 through pidchecked-1).
426 nextpid = nextpid + 1 + random_offset;
429 * If the process ID prototype has wrapped around,
430 * restart somewhat above 0, as the low-numbered procs
431 * tend to include daemons that don't exit.
433 if (nextpid >= PID_MAX) {
434 nextpid = nextpid % PID_MAX;
439 if (nextpid >= pidchecked) {
442 pidchecked = PID_MAX;
444 * Scan the active and zombie procs to check whether this pid
445 * is in use. Remember the lowest pid that's greater
446 * than nextpid, so we can avoid checking for a while.
448 p = LIST_FIRST(&allproc);
450 for (; p != 0; p = LIST_NEXT(p, p_list)) {
451 while (p->p_pid == nextpid ||
452 p->p_pgrp->pg_id == nextpid ||
453 p->p_session->s_sid == nextpid) {
455 if (nextpid >= pidchecked)
458 if (p->p_pid > nextpid && pidchecked > p->p_pid)
459 pidchecked = p->p_pid;
460 if (p->p_pgrp->pg_id > nextpid &&
461 pidchecked > p->p_pgrp->pg_id)
462 pidchecked = p->p_pgrp->pg_id;
463 if (p->p_session->s_sid > nextpid &&
464 pidchecked > p->p_session->s_sid)
465 pidchecked = p->p_session->s_sid;
469 p = LIST_FIRST(&zombproc);
477 * Called from exit1 to remove a process from the allproc
478 * list and move it to the zombie list.
483 proc_move_allproc_zombie(struct proc *p)
485 spin_lock_wr(&allproc_spin);
487 spin_unlock_wr(&allproc_spin);
488 tsleep(p, 0, "reap1", hz / 10);
489 spin_lock_wr(&allproc_spin);
491 LIST_REMOVE(p, p_list);
492 LIST_INSERT_HEAD(&zombproc, p, p_list);
493 LIST_REMOVE(p, p_hash);
495 spin_unlock_wr(&allproc_spin);
500 * This routine is called from kern_wait() and will remove the process
501 * from the zombie list and the sibling list. This routine will block
502 * if someone has a lock on the proces (p_lock).
507 proc_remove_zombie(struct proc *p)
509 spin_lock_wr(&allproc_spin);
511 spin_unlock_wr(&allproc_spin);
512 tsleep(p, 0, "reap1", hz / 10);
513 spin_lock_wr(&allproc_spin);
515 LIST_REMOVE(p, p_list); /* off zombproc */
516 LIST_REMOVE(p, p_sibling);
517 spin_unlock_wr(&allproc_spin);
521 * Scan all processes on the allproc list. The process is automatically
522 * held for the callback. A return value of -1 terminates the loop.
527 allproc_scan(int (*callback)(struct proc *, void *), void *data)
532 spin_lock_rd(&allproc_spin);
533 LIST_FOREACH(p, &allproc, p_list) {
535 spin_unlock_rd(&allproc_spin);
536 r = callback(p, data);
537 spin_lock_rd(&allproc_spin);
542 spin_unlock_rd(&allproc_spin);
546 * Scan all lwps of processes on the allproc list. The lwp is automatically
547 * held for the callback. A return value of -1 terminates the loop.
549 * possibly not MPSAFE, needs to access foreingn proc structures
552 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
558 spin_lock_rd(&allproc_spin);
559 LIST_FOREACH(p, &allproc, p_list) {
561 spin_unlock_rd(&allproc_spin);
562 FOREACH_LWP_IN_PROC(lp, p) {
564 r = callback(lp, data);
567 spin_lock_rd(&allproc_spin);
572 spin_unlock_rd(&allproc_spin);
576 * Scan all processes on the zombproc list. The process is automatically
577 * held for the callback. A return value of -1 terminates the loop.
582 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
587 spin_lock_rd(&allproc_spin);
588 LIST_FOREACH(p, &zombproc, p_list) {
590 spin_unlock_rd(&allproc_spin);
591 r = callback(p, data);
592 spin_lock_rd(&allproc_spin);
597 spin_unlock_rd(&allproc_spin);
604 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
610 for (i = 0; i <= pgrphash; i++) {
611 if (!LIST_EMPTY(&pgrphashtbl[i])) {
612 kprintf("\tindx %d\n", i);
613 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
615 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
616 (void *)pgrp, (long)pgrp->pg_id,
617 (void *)pgrp->pg_session,
618 pgrp->pg_session->s_count,
619 (void *)LIST_FIRST(&pgrp->pg_members));
620 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
621 kprintf("\t\tpid %ld addr %p pgrp %p\n",
622 (long)p->p_pid, (void *)p,
632 * Locate a process on the zombie list. Return a held process or NULL.
639 LIST_FOREACH(p, &zombproc, p_list)
646 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
648 struct kinfo_proc ki;
650 int skp = 0, had_output = 0;
653 bzero(&ki, sizeof(ki));
654 fill_kinfo_proc(p, &ki);
655 if ((flags & KERN_PROC_FLAG_LWP) == 0)
658 FOREACH_LWP_IN_PROC(lp, p) {
660 fill_kinfo_lwp(lp, &ki.kp_lwp);
662 error = SYSCTL_OUT(req, &ki, sizeof(ki));
669 /* We need to output at least the proc, even if there is no lwp. */
670 if (had_output == 0) {
671 error = SYSCTL_OUT(req, &ki, sizeof(ki));
677 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
679 struct kinfo_proc ki;
682 fill_kinfo_proc_kthread(td, &ki);
683 error = SYSCTL_OUT(req, &ki, sizeof(ki));
690 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
692 int *name = (int*) arg1;
693 int oid = oidp->oid_number;
694 u_int namelen = arg2;
696 struct proclist *plist;
698 int doingzomb, flags = 0;
702 struct ucred *cr1 = curproc->p_ucred;
704 flags = oid & KERN_PROC_FLAGMASK;
705 oid &= ~KERN_PROC_FLAGMASK;
707 if ((oid == KERN_PROC_ALL && namelen != 0) ||
708 (oid != KERN_PROC_ALL && namelen != 1))
711 if (oid == KERN_PROC_PID) {
712 p = pfind((pid_t)name[0]);
715 if (!PRISON_CHECK(cr1, p->p_ucred))
718 error = sysctl_out_proc(p, req, flags);
724 /* overestimate by 5 procs */
725 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
729 for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
734 LIST_FOREACH(p, plist, p_list) {
736 * Show a user only their processes.
738 if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
741 * Skip embryonic processes.
743 if (p->p_stat == SIDL)
746 * TODO - make more efficient (see notes below).
751 /* could do this by traversing pgrp */
752 if (p->p_pgrp == NULL ||
753 p->p_pgrp->pg_id != (pid_t)name[0])
758 if ((p->p_flag & P_CONTROLT) == 0 ||
759 p->p_session == NULL ||
760 p->p_session->s_ttyp == NULL ||
761 dev2udev(p->p_session->s_ttyp->t_dev) !=
767 if (p->p_ucred == NULL ||
768 p->p_ucred->cr_uid != (uid_t)name[0])
773 if (p->p_ucred == NULL ||
774 p->p_ucred->cr_ruid != (uid_t)name[0])
779 if (!PRISON_CHECK(cr1, p->p_ucred))
782 error = sysctl_out_proc(p, req, flags);
790 * Iterate over all active cpus and scan their thread list. Start
791 * with the next logical cpu and end with our original cpu. We
792 * migrate our own thread to each target cpu in order to safely scan
793 * its thread list. In the last loop we migrate back to our original
796 origcpu = mycpu->gd_cpuid;
797 if (!ps_showallthreads || jailed(cr1))
799 for (n = 1; n <= ncpus; ++n) {
803 nid = (origcpu + n) % ncpus;
804 if ((smp_active_mask & (1 << nid)) == 0)
806 rgd = globaldata_find(nid);
807 lwkt_setcpu_self(rgd);
809 TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
822 error = sysctl_out_proc_kthread(td, req, doingzomb);
833 * This sysctl allows a process to retrieve the argument list or process
834 * title for another process without groping around in the address space
835 * of the other process. It also allow a process to set its own "process
836 * title to a string of its own choice.
839 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
841 int *name = (int*) arg1;
842 u_int namelen = arg2;
846 struct ucred *cr1 = curproc->p_ucred;
851 p = pfind((pid_t)name[0]);
855 if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
858 if (req->newptr && curproc != p)
861 if (req->oldptr && p->p_args != NULL)
862 error = SYSCTL_OUT(req, p->p_args->ar_args, p->p_args->ar_length);
863 if (req->newptr == NULL)
866 if (p->p_args && --p->p_args->ar_ref == 0)
867 FREE(p->p_args, M_PARGS);
870 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
873 MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
876 pa->ar_length = req->newlen;
877 error = SYSCTL_IN(req, pa->ar_args, req->newlen);
885 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table");
887 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
888 0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
890 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
891 sysctl_kern_proc, "Process table");
893 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
894 sysctl_kern_proc, "Process table");
896 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
897 sysctl_kern_proc, "Process table");
899 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
900 sysctl_kern_proc, "Process table");
902 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
903 sysctl_kern_proc, "Process table");
905 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
906 sysctl_kern_proc, "Process table");
908 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
909 sysctl_kern_proc, "Process table");
911 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
912 sysctl_kern_proc, "Process table");
914 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
915 sysctl_kern_proc, "Process table");
917 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
918 sysctl_kern_proc, "Process table");
920 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
921 sysctl_kern_proc, "Process table");
923 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
924 sysctl_kern_proc_args, "Process argument list");